TasksSotADatasetsPapersMethodsSubmitAbout
Papers With Code 2

A community resource for machine learning research: papers, code, benchmarks, and state-of-the-art results.

Explore

Notable BenchmarksAll SotADatasetsPapersMethods

Community

Submit ResultsAbout

Data sourced from the PWC Archive (CC-BY-SA 4.0). Built by the community, for the community.

Papers/Cascade RPN: Delving into High-Quality Region Proposal Net...

Cascade RPN: Delving into High-Quality Region Proposal Network with Adaptive Convolution

Thang Vu, Hyunjun Jang, Trung X. Pham, Chang D. Yoo

2019-09-15NeurIPS 2019 12Region ProposalObject Detection
PaperPDFCodeCode(official)

Abstract

This paper considers an architecture referred to as Cascade Region Proposal Network (Cascade RPN) for improving the region-proposal quality and detection performance by \textit{systematically} addressing the limitation of the conventional RPN that \textit{heuristically defines} the anchors and \textit{aligns} the features to the anchors. First, instead of using multiple anchors with predefined scales and aspect ratios, Cascade RPN relies on a \textit{single anchor} per location and performs multi-stage refinement. Each stage is progressively more stringent in defining positive samples by starting out with an anchor-free metric followed by anchor-based metrics in the ensuing stages. Second, to attain alignment between the features and the anchors throughout the stages, \textit{adaptive convolution} is proposed that takes the anchors in addition to the image features as its input and learns the sampled features guided by the anchors. A simple implementation of a two-stage Cascade RPN achieves AR 13.4 points higher than that of the conventional RPN, surpassing any existing region proposal methods. When adopting to Fast R-CNN and Faster R-CNN, Cascade RPN can improve the detection mAP by 3.1 and 3.5 points, respectively. The code is made publicly available at \url{https://github.com/thangvubk/Cascade-RPN.git}.

Results

TaskDatasetMetricValueModel
Object DetectionCOCO test-devAP5058.9Faster R-CNN (Cascade RPN)
Object DetectionCOCO test-devAP7544.5Faster R-CNN (Cascade RPN)
Object DetectionCOCO test-devAPL52.6Faster R-CNN (Cascade RPN)
Object DetectionCOCO test-devAPM42.8Faster R-CNN (Cascade RPN)
Object DetectionCOCO test-devAPS22Faster R-CNN (Cascade RPN)
Object DetectionCOCO test-devbox mAP40.6Faster R-CNN (Cascade RPN)
Object DetectionCOCO test-devAP5059.4Fast R-CNN (Cascade RPN)
Object DetectionCOCO test-devAP7543.8Fast R-CNN (Cascade RPN)
Object DetectionCOCO test-devAPL51.6Fast R-CNN (Cascade RPN)
Object DetectionCOCO test-devAPM42.4Fast R-CNN (Cascade RPN)
Object DetectionCOCO test-devAPS22.1Fast R-CNN (Cascade RPN)
Object DetectionCOCO test-devbox mAP40.1Fast R-CNN (Cascade RPN)
3DCOCO test-devAP5058.9Faster R-CNN (Cascade RPN)
3DCOCO test-devAP7544.5Faster R-CNN (Cascade RPN)
3DCOCO test-devAPL52.6Faster R-CNN (Cascade RPN)
3DCOCO test-devAPM42.8Faster R-CNN (Cascade RPN)
3DCOCO test-devAPS22Faster R-CNN (Cascade RPN)
3DCOCO test-devbox mAP40.6Faster R-CNN (Cascade RPN)
3DCOCO test-devAP5059.4Fast R-CNN (Cascade RPN)
3DCOCO test-devAP7543.8Fast R-CNN (Cascade RPN)
3DCOCO test-devAPL51.6Fast R-CNN (Cascade RPN)
3DCOCO test-devAPM42.4Fast R-CNN (Cascade RPN)
3DCOCO test-devAPS22.1Fast R-CNN (Cascade RPN)
3DCOCO test-devbox mAP40.1Fast R-CNN (Cascade RPN)
2D ClassificationCOCO test-devAP5058.9Faster R-CNN (Cascade RPN)
2D ClassificationCOCO test-devAP7544.5Faster R-CNN (Cascade RPN)
2D ClassificationCOCO test-devAPL52.6Faster R-CNN (Cascade RPN)
2D ClassificationCOCO test-devAPM42.8Faster R-CNN (Cascade RPN)
2D ClassificationCOCO test-devAPS22Faster R-CNN (Cascade RPN)
2D ClassificationCOCO test-devbox mAP40.6Faster R-CNN (Cascade RPN)
2D ClassificationCOCO test-devAP5059.4Fast R-CNN (Cascade RPN)
2D ClassificationCOCO test-devAP7543.8Fast R-CNN (Cascade RPN)
2D ClassificationCOCO test-devAPL51.6Fast R-CNN (Cascade RPN)
2D ClassificationCOCO test-devAPM42.4Fast R-CNN (Cascade RPN)
2D ClassificationCOCO test-devAPS22.1Fast R-CNN (Cascade RPN)
2D ClassificationCOCO test-devbox mAP40.1Fast R-CNN (Cascade RPN)
2D Object DetectionCOCO test-devAP5058.9Faster R-CNN (Cascade RPN)
2D Object DetectionCOCO test-devAP7544.5Faster R-CNN (Cascade RPN)
2D Object DetectionCOCO test-devAPL52.6Faster R-CNN (Cascade RPN)
2D Object DetectionCOCO test-devAPM42.8Faster R-CNN (Cascade RPN)
2D Object DetectionCOCO test-devAPS22Faster R-CNN (Cascade RPN)
2D Object DetectionCOCO test-devbox mAP40.6Faster R-CNN (Cascade RPN)
2D Object DetectionCOCO test-devAP5059.4Fast R-CNN (Cascade RPN)
2D Object DetectionCOCO test-devAP7543.8Fast R-CNN (Cascade RPN)
2D Object DetectionCOCO test-devAPL51.6Fast R-CNN (Cascade RPN)
2D Object DetectionCOCO test-devAPM42.4Fast R-CNN (Cascade RPN)
2D Object DetectionCOCO test-devAPS22.1Fast R-CNN (Cascade RPN)
2D Object DetectionCOCO test-devbox mAP40.1Fast R-CNN (Cascade RPN)
16kCOCO test-devAP5058.9Faster R-CNN (Cascade RPN)
16kCOCO test-devAP7544.5Faster R-CNN (Cascade RPN)
16kCOCO test-devAPL52.6Faster R-CNN (Cascade RPN)
16kCOCO test-devAPM42.8Faster R-CNN (Cascade RPN)
16kCOCO test-devAPS22Faster R-CNN (Cascade RPN)
16kCOCO test-devbox mAP40.6Faster R-CNN (Cascade RPN)
16kCOCO test-devAP5059.4Fast R-CNN (Cascade RPN)
16kCOCO test-devAP7543.8Fast R-CNN (Cascade RPN)
16kCOCO test-devAPL51.6Fast R-CNN (Cascade RPN)
16kCOCO test-devAPM42.4Fast R-CNN (Cascade RPN)
16kCOCO test-devAPS22.1Fast R-CNN (Cascade RPN)
16kCOCO test-devbox mAP40.1Fast R-CNN (Cascade RPN)

Related Papers

A Real-Time System for Egocentric Hand-Object Interaction Detection in Industrial Domains2025-07-17RS-TinyNet: Stage-wise Feature Fusion Network for Detecting Tiny Objects in Remote Sensing Images2025-07-17Decoupled PROB: Decoupled Query Initialization Tasks and Objectness-Class Learning for Open World Object Detection2025-07-17Dual LiDAR-Based Traffic Movement Count Estimation at a Signalized Intersection: Deployment, Data Collection, and Preliminary Analysis2025-07-17Vision-based Perception for Autonomous Vehicles in Obstacle Avoidance Scenarios2025-07-16Tomato Multi-Angle Multi-Pose Dataset for Fine-Grained Phenotyping2025-07-15ECORE: Energy-Conscious Optimized Routing for Deep Learning Models at the Edge2025-07-08Beyond One Shot, Beyond One Perspective: Cross-View and Long-Horizon Distillation for Better LiDAR Representations2025-07-07